Latest / Columbus' Morning News / Is A.I. a "Jurassic Park" scenario waiting to happen?
Transcript
- 0:00Speaker 1: Let's go to the Legacy Retirement Group dot com phone
- 0:02Speaker 1: line and say good morning to Mike Debuski, ABC News
- 0:05Speaker 1: Technology reporter. Boy, it was a bad day to be
- 0:08Speaker 1: Sam Altman. He's the CEO of OpenAI, which is the
- 0:12Speaker 1: parent company of chat GPT. Mike, of course, and there
- 0:15Speaker 1: were some incidents at his house.
- 0:17Speaker 2: What's going on? Yeah, that's right. So over the weekend
- 0:20Speaker 2: we saw two separate incidents play out at Sam Altman's
- 0:23Speaker 2: home in San Francisco. The first one is the most serious.
- 0:27Speaker 2: This was a firebombing incident where a man allegedly drove
- 0:30Speaker 2: from his home in Texas to San Francisco, specifically to
- 0:34Speaker 2: target Sam Altman, throwing a Molotov cocktail at the front
- 0:38Speaker 2: gate of his house. Luckily, no one was hurt in
- 0:41Speaker 2: that incident, but later about an hour later, police picked
- 0:44Speaker 2: him up at the headquarters of OpenAI. So it appears that,
- 0:48Speaker 2: you know, Sam Altman and his company were the target
- 0:52Speaker 2: of this attack, and according to new charges unsealed late yesterday,
- 0:57Speaker 2: that indeed does seem to be the case. According to
- 0:59Speaker 2: a thar parties, this person was a man named Daniel
- 1:02Speaker 2: Moreno Gama. He's twenty years old and according to writings
- 1:06Speaker 2: that authorities found on his person when they picked him up.
- 1:09Speaker 2: He appeared to harbor a lot of anti AI sentiment,
- 1:12Speaker 2: not just against Sam Altman in Open AI, but the
- 1:15Speaker 2: entire AI industry. He apparently had a list of investors
- 1:19Speaker 2: and board members and executives at these companies that he
- 1:22Speaker 2: was attempting to target. The second incident at sam Altman's
- 1:27Speaker 2: home appears to be a shooting incident. It's remained sort
- 1:29Speaker 2: of unclear whether Sam Altman was the target of that,
- 1:32Speaker 2: but it took place near his home. Two people are
- 1:34Speaker 2: under arrest in connection with that case, and that investigation
- 1:37Speaker 2: is ongoing now. Authorities in the first instance, the firebombing
- 1:41Speaker 2: instance say there is no connection between these two things.
- 1:44Speaker 2: But even still, you can imagine that this was a
- 1:46Speaker 2: very tense moment in the Altman house, or tense weekend,
- 1:50Speaker 2: I should say. An Open AI spokesperson responded to our
- 1:54Speaker 2: request for comment yesterday saying that there is no place
- 1:57Speaker 2: in our democracy for violence against anyone, regardless of the
- 2:00Speaker 2: AI lab they work at, or the side of the
- 2:02Speaker 2: debate they belong to.
- 2:03Speaker 1: Yeah, and that's absolutely correct. I mean, this guy really
- 2:06Speaker 1: must not have liked AI. I mean dude, was there
- 2:09Speaker 1: any information in his writings, like what did he cite
- 2:13Speaker 1: that was he had such a problem with.
- 2:15Speaker 2: Yeah, I mean, it's a variety of different things from
- 2:18Speaker 2: what we understand. Specifically, what these core documents have outlined
- 2:22Speaker 2: is like this intent to kill. He wants to lead
- 2:26Speaker 2: by example, and it's some very ugly stuff, guys, frankly.
- 2:29Speaker 2: But it's also worth mentioning that this does come against
- 2:32Speaker 2: the backdrop of sort of a growing anti AI sentiment
- 2:36Speaker 2: out there. It's not so long ago that we were
- 2:37Speaker 2: talking about this Queen of Pax survey that found that
- 2:40Speaker 2: eighty percent of people that they surveyed feel concerned about
- 2:44Speaker 2: the future of artificial intelligence in some way or another.
- 2:48Speaker 2: Fifty five percent of Americans, according to this survey, think
- 2:50Speaker 2: that AI will do more harm than good in their
- 2:53Speaker 2: day to day lives. And perhaps no surprise there, we've
- 2:57Speaker 2: seen this technology, you know, run rampant for a few
- 3:00Speaker 2: years now. Lots of talk about it potentially displacing people's jobs,
- 3:05Speaker 2: although it's not really happening at scale yet. There is
- 3:08Speaker 2: concerns about its being used to spread misinformation online and
- 3:12Speaker 2: data centers, which are the sort of physical hardware that
- 3:15Speaker 2: supports the II Revolution. Basically big warehouses full of computers.
- 3:19Speaker 2: They've been cropping up across the country and driving up
- 3:22Speaker 2: electricity prices at a time when people are really feeling
- 3:24Speaker 2: the pinch in their wallets. This person, Daniel Moreno Gama,
- 3:28Speaker 2: in this instance does appear to be to have gone
- 3:31Speaker 2: through a mental health crisis of some kind, unfortunately, but
- 3:34Speaker 2: even still worth mentioning that sort of wider context.
- 3:36Speaker 1: Yeah, well, I'm glad that there was no no injuries
- 3:39Speaker 1: or worse in these incidents there that everybody's okay, and
- 3:42Speaker 1: this guy is in custody. The other story, speaking by
- 3:46Speaker 1: the way to Mike Debusky, ABC News Technology reporter, is mythos.
- 3:51Speaker 1: This is quad? Is this Anthropics AI tool? What is mythos?
- 3:58Speaker 2: Yeah, so Anthropic call basically all of their AI tools Claude,
- 4:02Speaker 2: and then this specific version of it has been getting
- 4:05Speaker 2: a lot of attention in the last week. It's called
- 4:07Speaker 2: Claude Mythos. The flavor here is that it's all about coding,
- 4:11Speaker 2: specifically detecting flaws in computer code. So if you're a
- 4:15Speaker 2: big business you do a lot of business online, you
- 4:18Speaker 2: have basically scores of computer code that you need to
- 4:21Speaker 2: sort through to make sure that your digital infrastructure is secure.
- 4:25Speaker 2: We hear about these data breaches all the time. Oftentimes
- 4:28Speaker 2: that's a hacker exploiting a flaw in code, and you know,
- 4:32Speaker 2: when the hackers make an advancement, the safety teams have
- 4:35Speaker 2: to make those advancements, and vice versa. It's an ongoing problem.
- 4:39Speaker 2: What Claude Mythos purports to do is automate that process.
- 4:43Speaker 2: They're saying, Hey, instead of hiring teams of people to
- 4:45Speaker 2: comb through your code, you can give it to this
- 4:47Speaker 2: AI and it will detect all these flaws that humans
- 4:51Speaker 2: might not be privy to. That sounds great right on
- 4:54Speaker 2: its face, right. It sounds like a tool that people
- 4:56Speaker 2: could use to make their systems more secure. It's also
- 5:00Speaker 2: away for hackers to basically evaluate all these cool new
- 5:03Speaker 2: ways to break into nights right, stealing your email or
- 5:06Speaker 2: your bank account information and whatnot. That is why Anthropics
- 5:09Speaker 2: says it's keeping Claude Mythos to itself for now. Right,
- 5:13Speaker 2: they announced it, but they're not allowing anyone to use
- 5:15Speaker 2: it outside of a handful of companies, big companies like
- 5:18Speaker 2: Apple and Amazon and Microsoft. But even still, they say
- 5:22Speaker 2: they're really limiting the rollout because they think that in
- 5:25Speaker 2: the wrong hands, this could wreak some real havoc.
- 5:27Speaker 1: I heard some longer form reporting by you, Mike. He
- 5:31Speaker 1: used a Jurassic Park reference of you know, we didn't.
- 5:34Speaker 1: We didn't. We're too wrapped up if we could. We
- 5:37Speaker 1: didn't think about if we should explain that for a second.
- 5:40Speaker 2: Yes, so, doctor Ian Malcolm from from Dress Golder'm played
- 5:44Speaker 2: by Jeff Goldboom, a star turn one of many from
- 5:47Speaker 2: Jeff Goldbloom. He said famously in that movie, your scientists
- 5:50Speaker 2: were so preoccupied with whether or not they could they
- 5:53Speaker 2: didn't stop to think if they should. That is, you know,
- 5:56Speaker 2: something that could be applied here, right. They were concerned that,
- 5:59Speaker 2: you know, resurrecting the dinosaurs was this huge scientific breakthrough,
- 6:03Speaker 2: but they didn't think about the potential for you know,
- 6:05Speaker 2: velociraptors to escape and reak havoc on a park.
- 6:08Speaker 1: Yeah.
- 6:08Speaker 2: I think there are some parallels to the AI space here.
- 6:11Speaker 2: But let's just also like level set here for a moment.
- 6:14Speaker 2: No one has really played around with Claude Mythos yet,
- 6:16Speaker 2: right outside of this handful of companies. That means that,
- 6:20Speaker 2: you know, we kind of just have to take anthropic
- 6:22Speaker 2: at their word. When it comes to how advanced, how
- 6:26Speaker 2: powerful this technology is, even like, you know, the potential
- 6:30Speaker 2: for this AI to go wrong is in some ways
- 6:34Speaker 2: an advertisement for Anthropic. Right. Anthropic has been pitching itself
- 6:37Speaker 2: as the adult in the AI space, right, the real
- 6:40Speaker 2: grown up, the responsible one. They even walked away from
- 6:43Speaker 2: a very lucrative deal with the Pentagon over concerns about
- 6:46Speaker 2: how their technology was going to be used. Now they're saying,
- 6:49Speaker 2: it's like, hey, we're playing it safe here. We created
- 6:52Speaker 2: this really powerful thing. We're capable of doing this, but
- 6:54Speaker 2: we're not going to release it just yet. I think
- 6:56Speaker 2: that we should take it all with a little bit
- 6:58Speaker 2: of a grain of salt before we, you know, kind
- 7:00Speaker 2: of get really scared about the potential for this stuff
- 7:02Speaker 2: to become Skynet or Jurassic Park or anything like that,
- 7:06Speaker 2: just because nobody's actually gotten a chance to evaluate it
- 7:09Speaker 2: for themselves.